That certainly is puzzling, given the content. I mean as an atheist, I do find the content objectionable, but hardly violent or graphic - whatever "graphic" is supposed to mean in this context.
Seems to me, this is a place where AI could do a real service. Instead of showing a questionable image or video, blur it and ask AI to overlay it - like FB did in your case - but with clean description of the content.
That doesn't address the step 1 problem - which is correctly spotting problematic content. But it would address the step 2 problem - which is what you faced: namely whether to click or not.
So, for example, you might see the blurred image and a statement like this "Image of the US flag superimposed over an Eagle along with a short prayer. Click to unblur."
Obviously a good AI could do a better job than I did, but something like that should be doable.
Heck, you could run an experiment doing this for all images and get some interesting feedback.
As an amusing aside, suppose they did that for all the ads? Make it so people have to opt in to see individual ads. "Advertisement for toilet paper depicting cartoon bears. Click to unblur." I wonder if that would kill ad revenue. Or, alternatively, if the ad folks would work harder to create content that people would want to click?